應用模糊類神經法則之球桿平衡控制系統
Abstract
摘要
使用古典控制理論設計控制器,須先找出受控系統的數學模型。因此控制器的性能與受控系統是否能被精確的描述有密切的關係。想要實現智慧型控制,就必須精確的掌握系統特性。「學習」是達到智慧型控制的第一步,經由「學習」可以降低影響動態性能的不確定因素。
模糊控制理論利用語意資訊,可以將人類知識和經驗轉換成控制法則,具有較佳的強健性和容錯性。類神經網路仿人腦的平行處理方式,具有學習功能,可應用於系統辨識和估測。
結合模糊推論和類神經網路之模糊類神經控制理論,則同時包含了類神經網路不受模式限定的學習能力,並可根據模糊邏輯,以萃取方式建構知識,具有補足類神經網路“黑盒子”缺點的能力。本研究即是利用模糊類神經法則之推理、學習之特性,並搭配手動操作裝置,將無法清楚描述的控制行為,以模糊推論方式轉換成語意式的模糊規則,並結合類神經網路之學習能力,以期能建立一個使用較少規則和數學,可以吸收人類知識並具有學習功能之球-桿平衡控制系統。
關鍵字:模糊推論、模糊類神經法則、球-桿平衡控制系統
Abstract It is necessary to find the mathematic model of the plant when we design the controller by classical control theory. Hence, the controller’s control ability is related to the plant which can be described accurately or not. If we want to make up intelligent control, it is necessary to get the system’s characteristics. Learning is the first step to achieve intelligent control. From learning, it can reduce the uncertain factor which can influence the dynamic system. Fuzzy control theory uses linguistic information, and it can transform human being’s knowledge and experiments to control rules. It has the better robustness and fault tolerance. Artificial neural network mimics a human brain’s parallel calculation. It has learning capability and it can be applied to system identification and estimate. The control theory combining with fuzzy reasoning system and artificial neural network not only have neural network’s learning capability, but also can build knowledge by extracting information form fuzzy logic. Hence it makes up neural network’s drawback which are always treated like a “black box”. This study utilizes the reasoning and learning ability of Fuzzy-Neural rules, and we will use the Fuzzy inference method to transfer the control behavior which can not be expressed clearly to linguistic Fuzzy rule with manual operation device. We will combine they with Neural Networks to establish a ball-beam balance control system which could assimilate human expertise with less rule and mathematics, and learning capability. Keyword: Fuzzy Inference, Fuzzy-Neural Rule, Ball-Beam Balance Control System.
Abstract It is necessary to find the mathematic model of the plant when we design the controller by classical control theory. Hence, the controller’s control ability is related to the plant which can be described accurately or not. If we want to make up intelligent control, it is necessary to get the system’s characteristics. Learning is the first step to achieve intelligent control. From learning, it can reduce the uncertain factor which can influence the dynamic system. Fuzzy control theory uses linguistic information, and it can transform human being’s knowledge and experiments to control rules. It has the better robustness and fault tolerance. Artificial neural network mimics a human brain’s parallel calculation. It has learning capability and it can be applied to system identification and estimate. The control theory combining with fuzzy reasoning system and artificial neural network not only have neural network’s learning capability, but also can build knowledge by extracting information form fuzzy logic. Hence it makes up neural network’s drawback which are always treated like a “black box”. This study utilizes the reasoning and learning ability of Fuzzy-Neural rules, and we will use the Fuzzy inference method to transfer the control behavior which can not be expressed clearly to linguistic Fuzzy rule with manual operation device. We will combine they with Neural Networks to establish a ball-beam balance control system which could assimilate human expertise with less rule and mathematics, and learning capability. Keyword: Fuzzy Inference, Fuzzy-Neural Rule, Ball-Beam Balance Control System.
Description
Keywords
模糊推論, 模糊類神經法則, 球-桿平衡控制系統, Fuzzy Inference, Fuzzy-Neural Rule, Ball-Beam Balance Control System